Page 216 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

Chapter 8
Survival curves of individual studies were obtained with Kaplan-Meier analyses for PFS. We used multivariable Cox proportional hazards models and multilevel Cox regression models to study the effects of timing and PET response criteria on PFS, TTP, and OS. Multilevel analyses were used to account for clustering of data within studies. To adjust for different inclusion criteria applied in the original studies, survival curves were corrected for IPI score. Corresponding hazard ratios (HRs) and their 95% confidence intervals (CIs) were obtained by Cox regression. For each variable included in the Cox regression model (timing I-PET, PET response criteria, and IPI score), the assumption of proportional hazards was assessed on the basis of Schoenfeld residuals [19], which was not violated. Univariable HRs were calculated for the DS4-5 response criterion and IPI score. To compare the discriminative ability of IPI score (low and low-intermediate vs high-intermediate and high) and age-adjusted IPI score (aaIPI; low and low- intermediate vs high-intermediate and high), univariable HRs of both prognostic scores in patients age 60 years or younger were calculated.
Diagnostic measures (positive predictive value [PPV] and negative predictive value [NPV]) were estimated from the Cox regression model probabilities of the event outcome (PPV) or survival probabilities (NPV) stratified for I-PET timing for DS4-5, DS5, and ∆SUVmax response criteria on 2-year PFS, TTP, and OS. Statistical analysis was performed using IBM SPSS version 24 and R version 3.6.3. A P value of < .05 was considered statistically significant.
Results
Patients’ characteristics
There were 2122 treatment-naïve DLBCL patients in the PETRA database, and 1692 of them were included in this IPD analysis (Figure 1). Patients who were treated with regimens other than R-CHOP (n = 107), who were ineligible for the original study (n = 101), or who had an I-PET after 5 cycles (n = 11) were ineligible for this study. To avoid duplication, we excluded the Bologna patients from the IAEA study (n = 40). Other reasons for exclusion were missing I-PET results (n = 99), survival data (n = 38), or clinical data (n = 32) and were younger than age 18 years (n = 2). Descriptive statistics for the main patient and I-PET outcome variables are presented in supplemental Table 1. There was low risk of bias for individual studies
214